Skip to content

Latest commit

 

History

History
133 lines (86 loc) · 7.18 KB

README.md

File metadata and controls

133 lines (86 loc) · 7.18 KB
Raw

IRTEX-1.0

This is the working Scientific project about an Image Retrieval Engine

The project currently consists of 5 branches:

  1. master - production code ready
  2. develop - development merge of all feature branches
  3. server - feature branch for web server in Django
  4. retrieval-engine - feature branch for image retrieval engine using ML/DL/IR approaches
  5. milestone2 - deliverable branch for milestone 2.

The application can be initiated in the following manner but it has some prerequisites.

Python (Use Python 3) virtualenv must be installed in the system. More details on this link about installation and usage -> https://packaging.python.org/guides/installing-using-pip-and-virtual-environments/

C++ Compiler is needed to run Mahotas library. You can download it from this link. Just download and run it -> https://download.microsoft.com/download/5/f/7/5f7acaeb-8363-451f-9425-68a90f98b238/visualcppbuildtools_full.exe?fixForIE=.exe.

  • Clone the repository in the virtualenv directory created and checkout to milestone2 branch. (For example the virtualenv directory name is venv)

    cd venv

    On Linux and MacOS

    source bin/activate

    On Windows

    Scripts\activate.bat

    The git commands

    git clone https://github.com/tarunlnmiit/irtex-1.0.git

    git checkout milestone2

    git pull origin milestone2

  • Run the following command to install the dependencies.

    pip install -r requirements.txt

  • Finally the server can be started using the following command

    python3 manage.py runserver

    The server will be started at http://127.0.0.1:8000

The code on server currently consists of Color Layout Descriptor(CLD) generator and Region Based Shape Descriptor(RBSD) generator. The results sent to the frontend are currently ranked on the average similarities computed on CLD and RBSD along with the individual ranked results of CLD and RBSD. But, all the feature extraction implementations can be individually run as standalone python scripts as follows:

  • In order to run the CLD extractor file, the command is as follows: First navigate to directory color_layout_descriptor. There are 3 inputs needed for this file to execute.

    1. --path - input images path
    2. --output - path for features output
    3. --type - features to be output in csv or a pkl (pickle) file

    python3 CLDescriptor.py --path <dataset path> --output <output path> --type <feature output type csv or pkl>

  • In order to run the CLD extractor file on toy dataset, the command is as follows: First navigate to directory color_layout_descriptor. There is 1 input needed for this file to execute.

    1. --test-image - query image path

    python3 cld_test.py --test_image <query image path>

    If no query image is provided then one image from toy dataset is taken as query.

  • In order to run the RBSD extractor file, the command is as follows: First navigate to directory region_based_descriptor. There are 3 inputs needed for this file to execute.

    1. --path - input images path
    2. --output - path for features output
    3. --type - features to be output in csv or a pkl (pickle) file

    python3 MomentsExtractor.py --path <dataset path> --output <output path> --type <feature output type csv or pkl>

  • In order to run the RBSD extractor file on toy dataset, the command is as follows: First navigate to directory region_based_descriptor. There is 1 input needed for this file to execute.

    1. --path - query image path

    python3 rbsd_test.py --path <query image path>

    If no query image is provided then one image from toy dataset is taken as query.

  • In order to run the Segmenation extractor file on cifar10 dataset, the command is as follows: First navigate to directory segmentaion. There are 2 inputs needed for this file to execute.

    1. --path - query image path which is required
    2. --check - Number of images per class to compare the query with by default its consider 2 (optional)
    3. --method - Method for segmentation 0 = Threshold, 1 = SLIC By default 0 is selected

    python3 segmentation.py --path <query image path> --check <# images> --method <0/1>

  • In order to run the ResNet20(High level feature) extractor file on cifar10 dataset, the command is as follows: First navigate to directory resnet20_cifar_10_features. There is 1 input needed for this file to execute.

    1. --stop_at - number of items per folder to extract (optional) defaults to all

    python3 ResNet20FeatureExtractor.py --stop_at <# folder>

  • In order to run the ResNet20(High level feature) extractor file on toy dataset, the command is as follows: First navigate to directory resnet20_cifar_10_features. If no query image is provided, it takes, one image from toy_data as query. There is 1 input needed for this file to execute.

    1. --test_image - query image path (optional)

    python3 resnet_test.py --test_image <query image path>

  • In order to run the VGG16ImageNet(High level feature) extractor file on cifar10 dataset, the command is as follows: First navigate to directory vgg16_imagenet_features. There is 1 input needed for this file to execute.

    1. --stop_at - number of items per folder to extract (optional) defaults to all

    python3 VGG16FeatureExractor.py --stop_at <# folder>

  • In order to run the VGG16ImageNet(High level feature) extractor file on toy dataset, the command is as follows: First navigate to directory vgg16_imagenet_features. If no query image is provided, it takes, one image from toy_data as query. There is 1 input needed for this file to execute.

    1. --test_image - query image path (optional)

    python3 vgg_test.py --test_image <query image path>

  • Since the csv file generated are big in size and we are internally using pickle to read the extracted feature vectors, we have uploaded the pre-generated csv files on Google Drive.

    RBSD feature csv

    CLD feature csv

The above mentioned steps were for standalone execution of feature extraction scripts but it is also possible to hit the endpoints on the local server to fetch results based on features. They are as follows (A REST client like POSTMAN is best for such endpoints execution). A detailed excel sheet with REQUEST TYPE, REQUEST BODY, ENDPOINT, RESPONSE is available here.

The flow of the whole web application from a user's perspective is given in the repository of frontend. Here is the link for the same. https://github.com/tarunlnmiit/irtex-client